xref: /freebsd/sys/netpfil/ipfw/ip_fw2.c (revision 23090366f729c56cab62de74c7a51792357e98a9)
1 /*-
2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3  *
4  * Redistribution and use in source and binary forms, with or without
5  * modification, are permitted provided that the following conditions
6  * are met:
7  * 1. Redistributions of source code must retain the above copyright
8  *    notice, this list of conditions and the following disclaimer.
9  * 2. Redistributions in binary form must reproduce the above copyright
10  *    notice, this list of conditions and the following disclaimer in the
11  *    documentation and/or other materials provided with the distribution.
12  *
13  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23  * SUCH DAMAGE.
24  */
25 
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
28 
29 /*
30  * The FreeBSD IP packet firewall, main file
31  */
32 
33 #include "opt_ipfw.h"
34 #include "opt_ipdivert.h"
35 #include "opt_inet.h"
36 #ifndef INET
37 #error "IPFIREWALL requires INET"
38 #endif /* INET */
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/condvar.h>
45 #include <sys/eventhandler.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/jail.h>
51 #include <sys/module.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/rwlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/syslog.h>
59 #include <sys/ucred.h>
60 #include <net/ethernet.h> /* for ETHERTYPE_IP */
61 #include <net/if.h>
62 #include <net/route.h>
63 #include <net/pf_mtag.h>
64 #include <net/pfil.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_icmp.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_carp.h>
75 #include <netinet/pim.h>
76 #include <netinet/tcp_var.h>
77 #include <netinet/udp.h>
78 #include <netinet/udp_var.h>
79 #include <netinet/sctp.h>
80 
81 #include <netinet/ip6.h>
82 #include <netinet/icmp6.h>
83 #ifdef INET6
84 #include <netinet6/in6_pcb.h>
85 #include <netinet6/scope6_var.h>
86 #include <netinet6/ip6_var.h>
87 #endif
88 
89 #include <netpfil/ipfw/ip_fw_private.h>
90 
91 #include <machine/in_cksum.h>	/* XXX for in_cksum */
92 
93 #ifdef MAC
94 #include <security/mac/mac_framework.h>
95 #endif
96 
97 /*
98  * static variables followed by global ones.
99  * All ipfw global variables are here.
100  */
101 
102 /* ipfw_vnet_ready controls when we are open for business */
103 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
104 #define	V_ipfw_vnet_ready	VNET(ipfw_vnet_ready)
105 
106 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
107 #define	V_fw_deny_unknown_exthdrs	VNET(fw_deny_unknown_exthdrs)
108 
109 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
110 #define	V_fw_permit_single_frag6	VNET(fw_permit_single_frag6)
111 
112 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
113 static int default_to_accept = 1;
114 #else
115 static int default_to_accept;
116 #endif
117 
118 VNET_DEFINE(int, autoinc_step);
119 VNET_DEFINE(int, fw_one_pass) = 1;
120 
121 VNET_DEFINE(unsigned int, fw_tables_max);
122 /* Use 128 tables by default */
123 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
124 
125 /*
126  * Each rule belongs to one of 32 different sets (0..31).
127  * The variable set_disable contains one bit per set.
128  * If the bit is set, all rules in the corresponding set
129  * are disabled. Set RESVD_SET(31) is reserved for the default rule
130  * and rules that are not deleted by the flush command,
131  * and CANNOT be disabled.
132  * Rules in set RESVD_SET can only be deleted individually.
133  */
134 VNET_DEFINE(u_int32_t, set_disable);
135 #define	V_set_disable			VNET(set_disable)
136 
137 VNET_DEFINE(int, fw_verbose);
138 /* counter for ipfw_log(NULL...) */
139 VNET_DEFINE(u_int64_t, norule_counter);
140 VNET_DEFINE(int, verbose_limit);
141 
142 /* layer3_chain contains the list of rules for layer 3 */
143 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
144 
145 ipfw_nat_t *ipfw_nat_ptr = NULL;
146 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
147 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
148 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
149 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
150 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
151 
152 #ifdef SYSCTL_NODE
153 uint32_t dummy_def = IPFW_DEFAULT_RULE;
154 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
155 
156 SYSBEGIN(f3)
157 
158 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
159 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
160     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
161     "Only do a single pass through ipfw when using dummynet(4)");
162 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
163     CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
164     "Rule number auto-increment step");
165 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
166     CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
167     "Log matches to ipfw rules");
168 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
169     CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
170     "Set upper limit of matches of ipfw rules logged");
171 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
172     &dummy_def, 0,
173     "The default/max possible rule number.");
174 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
175     CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
176     "Maximum number of tables");
177 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
178     &default_to_accept, 0,
179     "Make the default rule accept all packets.");
180 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
181 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
182 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
183     CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
184     "Number of static rules");
185 
186 #ifdef INET6
187 SYSCTL_DECL(_net_inet6_ip6);
188 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
189 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
190     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
191     "Deny packets with unknown IPv6 Extension Headers");
192 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
193     CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
194     "Permit single packet IPv6 fragments");
195 #endif /* INET6 */
196 
197 SYSEND
198 
199 #endif /* SYSCTL_NODE */
200 
201 
202 /*
203  * Some macros used in the various matching options.
204  * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
205  * Other macros just cast void * into the appropriate type
206  */
207 #define	L3HDR(T, ip)	((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
208 #define	TCP(p)		((struct tcphdr *)(p))
209 #define	SCTP(p)		((struct sctphdr *)(p))
210 #define	UDP(p)		((struct udphdr *)(p))
211 #define	ICMP(p)		((struct icmphdr *)(p))
212 #define	ICMP6(p)	((struct icmp6_hdr *)(p))
213 
214 static __inline int
215 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
216 {
217 	int type = icmp->icmp_type;
218 
219 	return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
220 }
221 
222 #define TT	( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
223     (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
224 
225 static int
226 is_icmp_query(struct icmphdr *icmp)
227 {
228 	int type = icmp->icmp_type;
229 
230 	return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
231 }
232 #undef TT
233 
234 /*
235  * The following checks use two arrays of 8 or 16 bits to store the
236  * bits that we want set or clear, respectively. They are in the
237  * low and high half of cmd->arg1 or cmd->d[0].
238  *
239  * We scan options and store the bits we find set. We succeed if
240  *
241  *	(want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
242  *
243  * The code is sometimes optimized not to store additional variables.
244  */
245 
246 static int
247 flags_match(ipfw_insn *cmd, u_int8_t bits)
248 {
249 	u_char want_clear;
250 	bits = ~bits;
251 
252 	if ( ((cmd->arg1 & 0xff) & bits) != 0)
253 		return 0; /* some bits we want set were clear */
254 	want_clear = (cmd->arg1 >> 8) & 0xff;
255 	if ( (want_clear & bits) != want_clear)
256 		return 0; /* some bits we want clear were set */
257 	return 1;
258 }
259 
260 static int
261 ipopts_match(struct ip *ip, ipfw_insn *cmd)
262 {
263 	int optlen, bits = 0;
264 	u_char *cp = (u_char *)(ip + 1);
265 	int x = (ip->ip_hl << 2) - sizeof (struct ip);
266 
267 	for (; x > 0; x -= optlen, cp += optlen) {
268 		int opt = cp[IPOPT_OPTVAL];
269 
270 		if (opt == IPOPT_EOL)
271 			break;
272 		if (opt == IPOPT_NOP)
273 			optlen = 1;
274 		else {
275 			optlen = cp[IPOPT_OLEN];
276 			if (optlen <= 0 || optlen > x)
277 				return 0; /* invalid or truncated */
278 		}
279 		switch (opt) {
280 
281 		default:
282 			break;
283 
284 		case IPOPT_LSRR:
285 			bits |= IP_FW_IPOPT_LSRR;
286 			break;
287 
288 		case IPOPT_SSRR:
289 			bits |= IP_FW_IPOPT_SSRR;
290 			break;
291 
292 		case IPOPT_RR:
293 			bits |= IP_FW_IPOPT_RR;
294 			break;
295 
296 		case IPOPT_TS:
297 			bits |= IP_FW_IPOPT_TS;
298 			break;
299 		}
300 	}
301 	return (flags_match(cmd, bits));
302 }
303 
304 static int
305 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
306 {
307 	int optlen, bits = 0;
308 	u_char *cp = (u_char *)(tcp + 1);
309 	int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
310 
311 	for (; x > 0; x -= optlen, cp += optlen) {
312 		int opt = cp[0];
313 		if (opt == TCPOPT_EOL)
314 			break;
315 		if (opt == TCPOPT_NOP)
316 			optlen = 1;
317 		else {
318 			optlen = cp[1];
319 			if (optlen <= 0)
320 				break;
321 		}
322 
323 		switch (opt) {
324 
325 		default:
326 			break;
327 
328 		case TCPOPT_MAXSEG:
329 			bits |= IP_FW_TCPOPT_MSS;
330 			break;
331 
332 		case TCPOPT_WINDOW:
333 			bits |= IP_FW_TCPOPT_WINDOW;
334 			break;
335 
336 		case TCPOPT_SACK_PERMITTED:
337 		case TCPOPT_SACK:
338 			bits |= IP_FW_TCPOPT_SACK;
339 			break;
340 
341 		case TCPOPT_TIMESTAMP:
342 			bits |= IP_FW_TCPOPT_TS;
343 			break;
344 
345 		}
346 	}
347 	return (flags_match(cmd, bits));
348 }
349 
350 static int
351 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
352 {
353 	if (ifp == NULL)	/* no iface with this packet, match fails */
354 		return 0;
355 	/* Check by name or by IP address */
356 	if (cmd->name[0] != '\0') { /* match by name */
357 		if (cmd->name[0] == '\1') /* use tablearg to match */
358 			return ipfw_lookup_table_extended(chain, cmd->p.glob,
359 				ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
360 		/* Check name */
361 		if (cmd->p.glob) {
362 			if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
363 				return(1);
364 		} else {
365 			if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
366 				return(1);
367 		}
368 	} else {
369 #ifdef __FreeBSD__	/* and OSX too ? */
370 		struct ifaddr *ia;
371 
372 		if_addr_rlock(ifp);
373 		TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
374 			if (ia->ifa_addr->sa_family != AF_INET)
375 				continue;
376 			if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
377 			    (ia->ifa_addr))->sin_addr.s_addr) {
378 				if_addr_runlock(ifp);
379 				return(1);	/* match */
380 			}
381 		}
382 		if_addr_runlock(ifp);
383 #endif /* __FreeBSD__ */
384 	}
385 	return(0);	/* no match, fail ... */
386 }
387 
388 /*
389  * The verify_path function checks if a route to the src exists and
390  * if it is reachable via ifp (when provided).
391  *
392  * The 'verrevpath' option checks that the interface that an IP packet
393  * arrives on is the same interface that traffic destined for the
394  * packet's source address would be routed out of.
395  * The 'versrcreach' option just checks that the source address is
396  * reachable via any route (except default) in the routing table.
397  * These two are a measure to block forged packets. This is also
398  * commonly known as "anti-spoofing" or Unicast Reverse Path
399  * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
400  * is purposely reminiscent of the Cisco IOS command,
401  *
402  *   ip verify unicast reverse-path
403  *   ip verify unicast source reachable-via any
404  *
405  * which implements the same functionality. But note that the syntax
406  * is misleading, and the check may be performed on all IP packets
407  * whether unicast, multicast, or broadcast.
408  */
409 static int
410 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
411 {
412 #ifndef __FreeBSD__
413 	return 0;
414 #else
415 	struct route ro;
416 	struct sockaddr_in *dst;
417 
418 	bzero(&ro, sizeof(ro));
419 
420 	dst = (struct sockaddr_in *)&(ro.ro_dst);
421 	dst->sin_family = AF_INET;
422 	dst->sin_len = sizeof(*dst);
423 	dst->sin_addr = src;
424 	in_rtalloc_ign(&ro, 0, fib);
425 
426 	if (ro.ro_rt == NULL)
427 		return 0;
428 
429 	/*
430 	 * If ifp is provided, check for equality with rtentry.
431 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
432 	 * in order to pass packets injected back by if_simloop():
433 	 * if useloopback == 1 routing entry (via lo0) for our own address
434 	 * may exist, so we need to handle routing assymetry.
435 	 */
436 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
437 		RTFREE(ro.ro_rt);
438 		return 0;
439 	}
440 
441 	/* if no ifp provided, check if rtentry is not default route */
442 	if (ifp == NULL &&
443 	     satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
444 		RTFREE(ro.ro_rt);
445 		return 0;
446 	}
447 
448 	/* or if this is a blackhole/reject route */
449 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
450 		RTFREE(ro.ro_rt);
451 		return 0;
452 	}
453 
454 	/* found valid route */
455 	RTFREE(ro.ro_rt);
456 	return 1;
457 #endif /* __FreeBSD__ */
458 }
459 
460 #ifdef INET6
461 /*
462  * ipv6 specific rules here...
463  */
464 static __inline int
465 icmp6type_match (int type, ipfw_insn_u32 *cmd)
466 {
467 	return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
468 }
469 
470 static int
471 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
472 {
473 	int i;
474 	for (i=0; i <= cmd->o.arg1; ++i )
475 		if (curr_flow == cmd->d[i] )
476 			return 1;
477 	return 0;
478 }
479 
480 /* support for IP6_*_ME opcodes */
481 static int
482 search_ip6_addr_net (struct in6_addr * ip6_addr)
483 {
484 	struct ifnet *mdc;
485 	struct ifaddr *mdc2;
486 	struct in6_ifaddr *fdm;
487 	struct in6_addr copia;
488 
489 	TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
490 		if_addr_rlock(mdc);
491 		TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
492 			if (mdc2->ifa_addr->sa_family == AF_INET6) {
493 				fdm = (struct in6_ifaddr *)mdc2;
494 				copia = fdm->ia_addr.sin6_addr;
495 				/* need for leaving scope_id in the sock_addr */
496 				in6_clearscope(&copia);
497 				if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
498 					if_addr_runlock(mdc);
499 					return 1;
500 				}
501 			}
502 		}
503 		if_addr_runlock(mdc);
504 	}
505 	return 0;
506 }
507 
508 static int
509 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
510 {
511 	struct route_in6 ro;
512 	struct sockaddr_in6 *dst;
513 
514 	bzero(&ro, sizeof(ro));
515 
516 	dst = (struct sockaddr_in6 * )&(ro.ro_dst);
517 	dst->sin6_family = AF_INET6;
518 	dst->sin6_len = sizeof(*dst);
519 	dst->sin6_addr = *src;
520 
521 	in6_rtalloc_ign(&ro, 0, fib);
522 	if (ro.ro_rt == NULL)
523 		return 0;
524 
525 	/*
526 	 * if ifp is provided, check for equality with rtentry
527 	 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
528 	 * to support the case of sending packets to an address of our own.
529 	 * (where the former interface is the first argument of if_simloop()
530 	 *  (=ifp), the latter is lo0)
531 	 */
532 	if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
533 		RTFREE(ro.ro_rt);
534 		return 0;
535 	}
536 
537 	/* if no ifp provided, check if rtentry is not default route */
538 	if (ifp == NULL &&
539 	    IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
540 		RTFREE(ro.ro_rt);
541 		return 0;
542 	}
543 
544 	/* or if this is a blackhole/reject route */
545 	if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
546 		RTFREE(ro.ro_rt);
547 		return 0;
548 	}
549 
550 	/* found valid route */
551 	RTFREE(ro.ro_rt);
552 	return 1;
553 
554 }
555 
556 static int
557 is_icmp6_query(int icmp6_type)
558 {
559 	if ((icmp6_type <= ICMP6_MAXTYPE) &&
560 	    (icmp6_type == ICMP6_ECHO_REQUEST ||
561 	    icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
562 	    icmp6_type == ICMP6_WRUREQUEST ||
563 	    icmp6_type == ICMP6_FQDN_QUERY ||
564 	    icmp6_type == ICMP6_NI_QUERY))
565 		return (1);
566 
567 	return (0);
568 }
569 
570 static void
571 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
572 {
573 	struct mbuf *m;
574 
575 	m = args->m;
576 	if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
577 		struct tcphdr *tcp;
578 		tcp = (struct tcphdr *)((char *)ip6 + hlen);
579 
580 		if ((tcp->th_flags & TH_RST) == 0) {
581 			struct mbuf *m0;
582 			m0 = ipfw_send_pkt(args->m, &(args->f_id),
583 			    ntohl(tcp->th_seq), ntohl(tcp->th_ack),
584 			    tcp->th_flags | TH_RST);
585 			if (m0 != NULL)
586 				ip6_output(m0, NULL, NULL, 0, NULL, NULL,
587 				    NULL);
588 		}
589 		FREE_PKT(m);
590 	} else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
591 #if 0
592 		/*
593 		 * Unlike above, the mbufs need to line up with the ip6 hdr,
594 		 * as the contents are read. We need to m_adj() the
595 		 * needed amount.
596 		 * The mbuf will however be thrown away so we can adjust it.
597 		 * Remember we did an m_pullup on it already so we
598 		 * can make some assumptions about contiguousness.
599 		 */
600 		if (args->L3offset)
601 			m_adj(m, args->L3offset);
602 #endif
603 		icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
604 	} else
605 		FREE_PKT(m);
606 
607 	args->m = NULL;
608 }
609 
610 #endif /* INET6 */
611 
612 
613 /*
614  * sends a reject message, consuming the mbuf passed as an argument.
615  */
616 static void
617 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
618 {
619 
620 #if 0
621 	/* XXX When ip is not guaranteed to be at mtod() we will
622 	 * need to account for this */
623 	 * The mbuf will however be thrown away so we can adjust it.
624 	 * Remember we did an m_pullup on it already so we
625 	 * can make some assumptions about contiguousness.
626 	 */
627 	if (args->L3offset)
628 		m_adj(m, args->L3offset);
629 #endif
630 	if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
631 		icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
632 	} else if (args->f_id.proto == IPPROTO_TCP) {
633 		struct tcphdr *const tcp =
634 		    L3HDR(struct tcphdr, mtod(args->m, struct ip *));
635 		if ( (tcp->th_flags & TH_RST) == 0) {
636 			struct mbuf *m;
637 			m = ipfw_send_pkt(args->m, &(args->f_id),
638 				ntohl(tcp->th_seq), ntohl(tcp->th_ack),
639 				tcp->th_flags | TH_RST);
640 			if (m != NULL)
641 				ip_output(m, NULL, NULL, 0, NULL, NULL);
642 		}
643 		FREE_PKT(args->m);
644 	} else
645 		FREE_PKT(args->m);
646 	args->m = NULL;
647 }
648 
649 /*
650  * Support for uid/gid/jail lookup. These tests are expensive
651  * (because we may need to look into the list of active sockets)
652  * so we cache the results. ugid_lookupp is 0 if we have not
653  * yet done a lookup, 1 if we succeeded, and -1 if we tried
654  * and failed. The function always returns the match value.
655  * We could actually spare the variable and use *uc, setting
656  * it to '(void *)check_uidgid if we have no info, NULL if
657  * we tried and failed, or any other value if successful.
658  */
659 static int
660 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
661     struct ucred **uc)
662 {
663 #ifndef __FreeBSD__
664 	/* XXX */
665 	return cred_check(insn, proto, oif,
666 	    dst_ip, dst_port, src_ip, src_port,
667 	    (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
668 #else  /* FreeBSD */
669 	struct in_addr src_ip, dst_ip;
670 	struct inpcbinfo *pi;
671 	struct ipfw_flow_id *id;
672 	struct inpcb *pcb, *inp;
673 	struct ifnet *oif;
674 	int lookupflags;
675 	int match;
676 
677 	id = &args->f_id;
678 	inp = args->inp;
679 	oif = args->oif;
680 
681 	/*
682 	 * Check to see if the UDP or TCP stack supplied us with
683 	 * the PCB. If so, rather then holding a lock and looking
684 	 * up the PCB, we can use the one that was supplied.
685 	 */
686 	if (inp && *ugid_lookupp == 0) {
687 		INP_LOCK_ASSERT(inp);
688 		if (inp->inp_socket != NULL) {
689 			*uc = crhold(inp->inp_cred);
690 			*ugid_lookupp = 1;
691 		} else
692 			*ugid_lookupp = -1;
693 	}
694 	/*
695 	 * If we have already been here and the packet has no
696 	 * PCB entry associated with it, then we can safely
697 	 * assume that this is a no match.
698 	 */
699 	if (*ugid_lookupp == -1)
700 		return (0);
701 	if (id->proto == IPPROTO_TCP) {
702 		lookupflags = 0;
703 		pi = &V_tcbinfo;
704 	} else if (id->proto == IPPROTO_UDP) {
705 		lookupflags = INPLOOKUP_WILDCARD;
706 		pi = &V_udbinfo;
707 	} else
708 		return 0;
709 	lookupflags |= INPLOOKUP_RLOCKPCB;
710 	match = 0;
711 	if (*ugid_lookupp == 0) {
712 		if (id->addr_type == 6) {
713 #ifdef INET6
714 			if (oif == NULL)
715 				pcb = in6_pcblookup_mbuf(pi,
716 				    &id->src_ip6, htons(id->src_port),
717 				    &id->dst_ip6, htons(id->dst_port),
718 				    lookupflags, oif, args->m);
719 			else
720 				pcb = in6_pcblookup_mbuf(pi,
721 				    &id->dst_ip6, htons(id->dst_port),
722 				    &id->src_ip6, htons(id->src_port),
723 				    lookupflags, oif, args->m);
724 #else
725 			*ugid_lookupp = -1;
726 			return (0);
727 #endif
728 		} else {
729 			src_ip.s_addr = htonl(id->src_ip);
730 			dst_ip.s_addr = htonl(id->dst_ip);
731 			if (oif == NULL)
732 				pcb = in_pcblookup_mbuf(pi,
733 				    src_ip, htons(id->src_port),
734 				    dst_ip, htons(id->dst_port),
735 				    lookupflags, oif, args->m);
736 			else
737 				pcb = in_pcblookup_mbuf(pi,
738 				    dst_ip, htons(id->dst_port),
739 				    src_ip, htons(id->src_port),
740 				    lookupflags, oif, args->m);
741 		}
742 		if (pcb != NULL) {
743 			INP_RLOCK_ASSERT(pcb);
744 			*uc = crhold(pcb->inp_cred);
745 			*ugid_lookupp = 1;
746 			INP_RUNLOCK(pcb);
747 		}
748 		if (*ugid_lookupp == 0) {
749 			/*
750 			 * We tried and failed, set the variable to -1
751 			 * so we will not try again on this packet.
752 			 */
753 			*ugid_lookupp = -1;
754 			return (0);
755 		}
756 	}
757 	if (insn->o.opcode == O_UID)
758 		match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
759 	else if (insn->o.opcode == O_GID)
760 		match = groupmember((gid_t)insn->d[0], *uc);
761 	else if (insn->o.opcode == O_JAIL)
762 		match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
763 	return (match);
764 #endif /* __FreeBSD__ */
765 }
766 
767 /*
768  * Helper function to set args with info on the rule after the matching
769  * one. slot is precise, whereas we guess rule_id as they are
770  * assigned sequentially.
771  */
772 static inline void
773 set_match(struct ip_fw_args *args, int slot,
774 	struct ip_fw_chain *chain)
775 {
776 	args->rule.chain_id = chain->id;
777 	args->rule.slot = slot + 1; /* we use 0 as a marker */
778 	args->rule.rule_id = 1 + chain->map[slot]->id;
779 	args->rule.rulenum = chain->map[slot]->rulenum;
780 }
781 
782 /*
783  * The main check routine for the firewall.
784  *
785  * All arguments are in args so we can modify them and return them
786  * back to the caller.
787  *
788  * Parameters:
789  *
790  *	args->m	(in/out) The packet; we set to NULL when/if we nuke it.
791  *		Starts with the IP header.
792  *	args->eh (in)	Mac header if present, NULL for layer3 packet.
793  *	args->L3offset	Number of bytes bypassed if we came from L2.
794  *			e.g. often sizeof(eh)  ** NOTYET **
795  *	args->oif	Outgoing interface, NULL if packet is incoming.
796  *		The incoming interface is in the mbuf. (in)
797  *	args->divert_rule (in/out)
798  *		Skip up to the first rule past this rule number;
799  *		upon return, non-zero port number for divert or tee.
800  *
801  *	args->rule	Pointer to the last matching rule (in/out)
802  *	args->next_hop	Socket we are forwarding to (out).
803  *	args->next_hop6	IPv6 next hop we are forwarding to (out).
804  *	args->f_id	Addresses grabbed from the packet (out)
805  * 	args->rule.info	a cookie depending on rule action
806  *
807  * Return value:
808  *
809  *	IP_FW_PASS	the packet must be accepted
810  *	IP_FW_DENY	the packet must be dropped
811  *	IP_FW_DIVERT	divert packet, port in m_tag
812  *	IP_FW_TEE	tee packet, port in m_tag
813  *	IP_FW_DUMMYNET	to dummynet, pipe in args->cookie
814  *	IP_FW_NETGRAPH	into netgraph, cookie args->cookie
815  *		args->rule contains the matching rule,
816  *		args->rule.info has additional information.
817  *
818  */
819 int
820 ipfw_chk(struct ip_fw_args *args)
821 {
822 
823 	/*
824 	 * Local variables holding state while processing a packet:
825 	 *
826 	 * IMPORTANT NOTE: to speed up the processing of rules, there
827 	 * are some assumption on the values of the variables, which
828 	 * are documented here. Should you change them, please check
829 	 * the implementation of the various instructions to make sure
830 	 * that they still work.
831 	 *
832 	 * args->eh	The MAC header. It is non-null for a layer2
833 	 *	packet, it is NULL for a layer-3 packet.
834 	 * **notyet**
835 	 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
836 	 *
837 	 * m | args->m	Pointer to the mbuf, as received from the caller.
838 	 *	It may change if ipfw_chk() does an m_pullup, or if it
839 	 *	consumes the packet because it calls send_reject().
840 	 *	XXX This has to change, so that ipfw_chk() never modifies
841 	 *	or consumes the buffer.
842 	 * ip	is the beginning of the ip(4 or 6) header.
843 	 *	Calculated by adding the L3offset to the start of data.
844 	 *	(Until we start using L3offset, the packet is
845 	 *	supposed to start with the ip header).
846 	 */
847 	struct mbuf *m = args->m;
848 	struct ip *ip = mtod(m, struct ip *);
849 
850 	/*
851 	 * For rules which contain uid/gid or jail constraints, cache
852 	 * a copy of the users credentials after the pcb lookup has been
853 	 * executed. This will speed up the processing of rules with
854 	 * these types of constraints, as well as decrease contention
855 	 * on pcb related locks.
856 	 */
857 #ifndef __FreeBSD__
858 	struct bsd_ucred ucred_cache;
859 #else
860 	struct ucred *ucred_cache = NULL;
861 #endif
862 	int ucred_lookup = 0;
863 
864 	/*
865 	 * oif | args->oif	If NULL, ipfw_chk has been called on the
866 	 *	inbound path (ether_input, ip_input).
867 	 *	If non-NULL, ipfw_chk has been called on the outbound path
868 	 *	(ether_output, ip_output).
869 	 */
870 	struct ifnet *oif = args->oif;
871 
872 	int f_pos = 0;		/* index of current rule in the array */
873 	int retval = 0;
874 
875 	/*
876 	 * hlen	The length of the IP header.
877 	 */
878 	u_int hlen = 0;		/* hlen >0 means we have an IP pkt */
879 
880 	/*
881 	 * offset	The offset of a fragment. offset != 0 means that
882 	 *	we have a fragment at this offset of an IPv4 packet.
883 	 *	offset == 0 means that (if this is an IPv4 packet)
884 	 *	this is the first or only fragment.
885 	 *	For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
886 	 *	or there is a single packet fragement (fragement header added
887 	 *	without needed).  We will treat a single packet fragment as if
888 	 *	there was no fragment header (or log/block depending on the
889 	 *	V_fw_permit_single_frag6 sysctl setting).
890 	 */
891 	u_short offset = 0;
892 	u_short ip6f_mf = 0;
893 
894 	/*
895 	 * Local copies of addresses. They are only valid if we have
896 	 * an IP packet.
897 	 *
898 	 * proto	The protocol. Set to 0 for non-ip packets,
899 	 *	or to the protocol read from the packet otherwise.
900 	 *	proto != 0 means that we have an IPv4 packet.
901 	 *
902 	 * src_port, dst_port	port numbers, in HOST format. Only
903 	 *	valid for TCP and UDP packets.
904 	 *
905 	 * src_ip, dst_ip	ip addresses, in NETWORK format.
906 	 *	Only valid for IPv4 packets.
907 	 */
908 	uint8_t proto;
909 	uint16_t src_port = 0, dst_port = 0;	/* NOTE: host format	*/
910 	struct in_addr src_ip, dst_ip;		/* NOTE: network format	*/
911 	uint16_t iplen=0;
912 	int pktlen;
913 	uint16_t	etype = 0;	/* Host order stored ether type */
914 
915 	/*
916 	 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
917 	 * 	MATCH_NONE when checked and not matched (q = NULL),
918 	 *	MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
919 	 */
920 	int dyn_dir = MATCH_UNKNOWN;
921 	ipfw_dyn_rule *q = NULL;
922 	struct ip_fw_chain *chain = &V_layer3_chain;
923 
924 	/*
925 	 * We store in ulp a pointer to the upper layer protocol header.
926 	 * In the ipv4 case this is easy to determine from the header,
927 	 * but for ipv6 we might have some additional headers in the middle.
928 	 * ulp is NULL if not found.
929 	 */
930 	void *ulp = NULL;		/* upper layer protocol pointer. */
931 
932 	/* XXX ipv6 variables */
933 	int is_ipv6 = 0;
934 	uint8_t	icmp6_type = 0;
935 	uint16_t ext_hd = 0;	/* bits vector for extension header filtering */
936 	/* end of ipv6 variables */
937 
938 	int is_ipv4 = 0;
939 
940 	int done = 0;		/* flag to exit the outer loop */
941 
942 	if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
943 		return (IP_FW_PASS);	/* accept */
944 
945 	dst_ip.s_addr = 0;		/* make sure it is initialized */
946 	src_ip.s_addr = 0;		/* make sure it is initialized */
947 	pktlen = m->m_pkthdr.len;
948 	args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
949 	proto = args->f_id.proto = 0;	/* mark f_id invalid */
950 		/* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
951 
952 /*
953  * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
954  * then it sets p to point at the offset "len" in the mbuf. WARNING: the
955  * pointer might become stale after other pullups (but we never use it
956  * this way).
957  */
958 #define PULLUP_TO(_len, p, T)	PULLUP_LEN(_len, p, sizeof(T))
959 #define PULLUP_LEN(_len, p, T)					\
960 do {								\
961 	int x = (_len) + T;					\
962 	if ((m)->m_len < x) {					\
963 		args->m = m = m_pullup(m, x);			\
964 		if (m == NULL)					\
965 			goto pullup_failed;			\
966 	}							\
967 	p = (mtod(m, char *) + (_len));				\
968 } while (0)
969 
970 	/*
971 	 * if we have an ether header,
972 	 */
973 	if (args->eh)
974 		etype = ntohs(args->eh->ether_type);
975 
976 	/* Identify IP packets and fill up variables. */
977 	if (pktlen >= sizeof(struct ip6_hdr) &&
978 	    (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
979 		struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
980 		is_ipv6 = 1;
981 		args->f_id.addr_type = 6;
982 		hlen = sizeof(struct ip6_hdr);
983 		proto = ip6->ip6_nxt;
984 
985 		/* Search extension headers to find upper layer protocols */
986 		while (ulp == NULL && offset == 0) {
987 			switch (proto) {
988 			case IPPROTO_ICMPV6:
989 				PULLUP_TO(hlen, ulp, struct icmp6_hdr);
990 				icmp6_type = ICMP6(ulp)->icmp6_type;
991 				break;
992 
993 			case IPPROTO_TCP:
994 				PULLUP_TO(hlen, ulp, struct tcphdr);
995 				dst_port = TCP(ulp)->th_dport;
996 				src_port = TCP(ulp)->th_sport;
997 				/* save flags for dynamic rules */
998 				args->f_id._flags = TCP(ulp)->th_flags;
999 				break;
1000 
1001 			case IPPROTO_SCTP:
1002 				PULLUP_TO(hlen, ulp, struct sctphdr);
1003 				src_port = SCTP(ulp)->src_port;
1004 				dst_port = SCTP(ulp)->dest_port;
1005 				break;
1006 
1007 			case IPPROTO_UDP:
1008 				PULLUP_TO(hlen, ulp, struct udphdr);
1009 				dst_port = UDP(ulp)->uh_dport;
1010 				src_port = UDP(ulp)->uh_sport;
1011 				break;
1012 
1013 			case IPPROTO_HOPOPTS:	/* RFC 2460 */
1014 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1015 				ext_hd |= EXT_HOPOPTS;
1016 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1017 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1018 				ulp = NULL;
1019 				break;
1020 
1021 			case IPPROTO_ROUTING:	/* RFC 2460 */
1022 				PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1023 				switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1024 				case 0:
1025 					ext_hd |= EXT_RTHDR0;
1026 					break;
1027 				case 2:
1028 					ext_hd |= EXT_RTHDR2;
1029 					break;
1030 				default:
1031 					if (V_fw_verbose)
1032 						printf("IPFW2: IPV6 - Unknown "
1033 						    "Routing Header type(%d)\n",
1034 						    ((struct ip6_rthdr *)
1035 						    ulp)->ip6r_type);
1036 					if (V_fw_deny_unknown_exthdrs)
1037 					    return (IP_FW_DENY);
1038 					break;
1039 				}
1040 				ext_hd |= EXT_ROUTING;
1041 				hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1042 				proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1043 				ulp = NULL;
1044 				break;
1045 
1046 			case IPPROTO_FRAGMENT:	/* RFC 2460 */
1047 				PULLUP_TO(hlen, ulp, struct ip6_frag);
1048 				ext_hd |= EXT_FRAGMENT;
1049 				hlen += sizeof (struct ip6_frag);
1050 				proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1051 				offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1052 					IP6F_OFF_MASK;
1053 				ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1054 					IP6F_MORE_FRAG;
1055 				if (V_fw_permit_single_frag6 == 0 &&
1056 				    offset == 0 && ip6f_mf == 0) {
1057 					if (V_fw_verbose)
1058 						printf("IPFW2: IPV6 - Invalid "
1059 						    "Fragment Header\n");
1060 					if (V_fw_deny_unknown_exthdrs)
1061 					    return (IP_FW_DENY);
1062 					break;
1063 				}
1064 				args->f_id.extra =
1065 				    ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1066 				ulp = NULL;
1067 				break;
1068 
1069 			case IPPROTO_DSTOPTS:	/* RFC 2460 */
1070 				PULLUP_TO(hlen, ulp, struct ip6_hbh);
1071 				ext_hd |= EXT_DSTOPTS;
1072 				hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1073 				proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1074 				ulp = NULL;
1075 				break;
1076 
1077 			case IPPROTO_AH:	/* RFC 2402 */
1078 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1079 				ext_hd |= EXT_AH;
1080 				hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1081 				proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1082 				ulp = NULL;
1083 				break;
1084 
1085 			case IPPROTO_ESP:	/* RFC 2406 */
1086 				PULLUP_TO(hlen, ulp, uint32_t);	/* SPI, Seq# */
1087 				/* Anything past Seq# is variable length and
1088 				 * data past this ext. header is encrypted. */
1089 				ext_hd |= EXT_ESP;
1090 				break;
1091 
1092 			case IPPROTO_NONE:	/* RFC 2460 */
1093 				/*
1094 				 * Packet ends here, and IPv6 header has
1095 				 * already been pulled up. If ip6e_len!=0
1096 				 * then octets must be ignored.
1097 				 */
1098 				ulp = ip; /* non-NULL to get out of loop. */
1099 				break;
1100 
1101 			case IPPROTO_OSPFIGP:
1102 				/* XXX OSPF header check? */
1103 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1104 				break;
1105 
1106 			case IPPROTO_PIM:
1107 				/* XXX PIM header check? */
1108 				PULLUP_TO(hlen, ulp, struct pim);
1109 				break;
1110 
1111 			case IPPROTO_CARP:
1112 				PULLUP_TO(hlen, ulp, struct carp_header);
1113 				if (((struct carp_header *)ulp)->carp_version !=
1114 				    CARP_VERSION)
1115 					return (IP_FW_DENY);
1116 				if (((struct carp_header *)ulp)->carp_type !=
1117 				    CARP_ADVERTISEMENT)
1118 					return (IP_FW_DENY);
1119 				break;
1120 
1121 			case IPPROTO_IPV6:	/* RFC 2893 */
1122 				PULLUP_TO(hlen, ulp, struct ip6_hdr);
1123 				break;
1124 
1125 			case IPPROTO_IPV4:	/* RFC 2893 */
1126 				PULLUP_TO(hlen, ulp, struct ip);
1127 				break;
1128 
1129 			default:
1130 				if (V_fw_verbose)
1131 					printf("IPFW2: IPV6 - Unknown "
1132 					    "Extension Header(%d), ext_hd=%x\n",
1133 					     proto, ext_hd);
1134 				if (V_fw_deny_unknown_exthdrs)
1135 				    return (IP_FW_DENY);
1136 				PULLUP_TO(hlen, ulp, struct ip6_ext);
1137 				break;
1138 			} /*switch */
1139 		}
1140 		ip = mtod(m, struct ip *);
1141 		ip6 = (struct ip6_hdr *)ip;
1142 		args->f_id.src_ip6 = ip6->ip6_src;
1143 		args->f_id.dst_ip6 = ip6->ip6_dst;
1144 		args->f_id.src_ip = 0;
1145 		args->f_id.dst_ip = 0;
1146 		args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1147 	} else if (pktlen >= sizeof(struct ip) &&
1148 	    (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1149 	    	is_ipv4 = 1;
1150 		hlen = ip->ip_hl << 2;
1151 		args->f_id.addr_type = 4;
1152 
1153 		/*
1154 		 * Collect parameters into local variables for faster matching.
1155 		 */
1156 		proto = ip->ip_p;
1157 		src_ip = ip->ip_src;
1158 		dst_ip = ip->ip_dst;
1159 		offset = ntohs(ip->ip_off) & IP_OFFMASK;
1160 		iplen = ntohs(ip->ip_len);
1161 		pktlen = iplen < pktlen ? iplen : pktlen;
1162 
1163 		if (offset == 0) {
1164 			switch (proto) {
1165 			case IPPROTO_TCP:
1166 				PULLUP_TO(hlen, ulp, struct tcphdr);
1167 				dst_port = TCP(ulp)->th_dport;
1168 				src_port = TCP(ulp)->th_sport;
1169 				/* save flags for dynamic rules */
1170 				args->f_id._flags = TCP(ulp)->th_flags;
1171 				break;
1172 
1173 			case IPPROTO_SCTP:
1174 				PULLUP_TO(hlen, ulp, struct sctphdr);
1175 				src_port = SCTP(ulp)->src_port;
1176 				dst_port = SCTP(ulp)->dest_port;
1177 				break;
1178 
1179 			case IPPROTO_UDP:
1180 				PULLUP_TO(hlen, ulp, struct udphdr);
1181 				dst_port = UDP(ulp)->uh_dport;
1182 				src_port = UDP(ulp)->uh_sport;
1183 				break;
1184 
1185 			case IPPROTO_ICMP:
1186 				PULLUP_TO(hlen, ulp, struct icmphdr);
1187 				//args->f_id.flags = ICMP(ulp)->icmp_type;
1188 				break;
1189 
1190 			default:
1191 				break;
1192 			}
1193 		}
1194 
1195 		ip = mtod(m, struct ip *);
1196 		args->f_id.src_ip = ntohl(src_ip.s_addr);
1197 		args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1198 	}
1199 #undef PULLUP_TO
1200 	if (proto) { /* we may have port numbers, store them */
1201 		args->f_id.proto = proto;
1202 		args->f_id.src_port = src_port = ntohs(src_port);
1203 		args->f_id.dst_port = dst_port = ntohs(dst_port);
1204 	}
1205 
1206 	IPFW_RLOCK(chain);
1207 	if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1208 		IPFW_RUNLOCK(chain);
1209 		return (IP_FW_PASS);	/* accept */
1210 	}
1211 	if (args->rule.slot) {
1212 		/*
1213 		 * Packet has already been tagged as a result of a previous
1214 		 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1215 		 * REASS, NETGRAPH, DIVERT/TEE...)
1216 		 * Validate the slot and continue from the next one
1217 		 * if still present, otherwise do a lookup.
1218 		 */
1219 		f_pos = (args->rule.chain_id == chain->id) ?
1220 		    args->rule.slot :
1221 		    ipfw_find_rule(chain, args->rule.rulenum,
1222 			args->rule.rule_id);
1223 	} else {
1224 		f_pos = 0;
1225 	}
1226 
1227 	/*
1228 	 * Now scan the rules, and parse microinstructions for each rule.
1229 	 * We have two nested loops and an inner switch. Sometimes we
1230 	 * need to break out of one or both loops, or re-enter one of
1231 	 * the loops with updated variables. Loop variables are:
1232 	 *
1233 	 *	f_pos (outer loop) points to the current rule.
1234 	 *		On output it points to the matching rule.
1235 	 *	done (outer loop) is used as a flag to break the loop.
1236 	 *	l (inner loop)	residual length of current rule.
1237 	 *		cmd points to the current microinstruction.
1238 	 *
1239 	 * We break the inner loop by setting l=0 and possibly
1240 	 * cmdlen=0 if we don't want to advance cmd.
1241 	 * We break the outer loop by setting done=1
1242 	 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1243 	 * as needed.
1244 	 */
1245 	for (; f_pos < chain->n_rules; f_pos++) {
1246 		ipfw_insn *cmd;
1247 		uint32_t tablearg = 0;
1248 		int l, cmdlen, skip_or; /* skip rest of OR block */
1249 		struct ip_fw *f;
1250 
1251 		f = chain->map[f_pos];
1252 		if (V_set_disable & (1 << f->set) )
1253 			continue;
1254 
1255 		skip_or = 0;
1256 		for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1257 		    l -= cmdlen, cmd += cmdlen) {
1258 			int match;
1259 
1260 			/*
1261 			 * check_body is a jump target used when we find a
1262 			 * CHECK_STATE, and need to jump to the body of
1263 			 * the target rule.
1264 			 */
1265 
1266 /* check_body: */
1267 			cmdlen = F_LEN(cmd);
1268 			/*
1269 			 * An OR block (insn_1 || .. || insn_n) has the
1270 			 * F_OR bit set in all but the last instruction.
1271 			 * The first match will set "skip_or", and cause
1272 			 * the following instructions to be skipped until
1273 			 * past the one with the F_OR bit clear.
1274 			 */
1275 			if (skip_or) {		/* skip this instruction */
1276 				if ((cmd->len & F_OR) == 0)
1277 					skip_or = 0;	/* next one is good */
1278 				continue;
1279 			}
1280 			match = 0; /* set to 1 if we succeed */
1281 
1282 			switch (cmd->opcode) {
1283 			/*
1284 			 * The first set of opcodes compares the packet's
1285 			 * fields with some pattern, setting 'match' if a
1286 			 * match is found. At the end of the loop there is
1287 			 * logic to deal with F_NOT and F_OR flags associated
1288 			 * with the opcode.
1289 			 */
1290 			case O_NOP:
1291 				match = 1;
1292 				break;
1293 
1294 			case O_FORWARD_MAC:
1295 				printf("ipfw: opcode %d unimplemented\n",
1296 				    cmd->opcode);
1297 				break;
1298 
1299 			case O_GID:
1300 			case O_UID:
1301 			case O_JAIL:
1302 				/*
1303 				 * We only check offset == 0 && proto != 0,
1304 				 * as this ensures that we have a
1305 				 * packet with the ports info.
1306 				 */
1307 				if (offset != 0)
1308 					break;
1309 				if (proto == IPPROTO_TCP ||
1310 				    proto == IPPROTO_UDP)
1311 					match = check_uidgid(
1312 						    (ipfw_insn_u32 *)cmd,
1313 						    args, &ucred_lookup,
1314 #ifdef __FreeBSD__
1315 						    &ucred_cache);
1316 #else
1317 						    (void *)&ucred_cache);
1318 #endif
1319 				break;
1320 
1321 			case O_RECV:
1322 				match = iface_match(m->m_pkthdr.rcvif,
1323 				    (ipfw_insn_if *)cmd, chain, &tablearg);
1324 				break;
1325 
1326 			case O_XMIT:
1327 				match = iface_match(oif, (ipfw_insn_if *)cmd,
1328 				    chain, &tablearg);
1329 				break;
1330 
1331 			case O_VIA:
1332 				match = iface_match(oif ? oif :
1333 				    m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1334 				    chain, &tablearg);
1335 				break;
1336 
1337 			case O_MACADDR2:
1338 				if (args->eh != NULL) {	/* have MAC header */
1339 					u_int32_t *want = (u_int32_t *)
1340 						((ipfw_insn_mac *)cmd)->addr;
1341 					u_int32_t *mask = (u_int32_t *)
1342 						((ipfw_insn_mac *)cmd)->mask;
1343 					u_int32_t *hdr = (u_int32_t *)args->eh;
1344 
1345 					match =
1346 					    ( want[0] == (hdr[0] & mask[0]) &&
1347 					      want[1] == (hdr[1] & mask[1]) &&
1348 					      want[2] == (hdr[2] & mask[2]) );
1349 				}
1350 				break;
1351 
1352 			case O_MAC_TYPE:
1353 				if (args->eh != NULL) {
1354 					u_int16_t *p =
1355 					    ((ipfw_insn_u16 *)cmd)->ports;
1356 					int i;
1357 
1358 					for (i = cmdlen - 1; !match && i>0;
1359 					    i--, p += 2)
1360 						match = (etype >= p[0] &&
1361 						    etype <= p[1]);
1362 				}
1363 				break;
1364 
1365 			case O_FRAG:
1366 				match = (offset != 0);
1367 				break;
1368 
1369 			case O_IN:	/* "out" is "not in" */
1370 				match = (oif == NULL);
1371 				break;
1372 
1373 			case O_LAYER2:
1374 				match = (args->eh != NULL);
1375 				break;
1376 
1377 			case O_DIVERTED:
1378 			    {
1379 				/* For diverted packets, args->rule.info
1380 				 * contains the divert port (in host format)
1381 				 * reason and direction.
1382 				 */
1383 				uint32_t i = args->rule.info;
1384 				match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1385 				    cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1386 			    }
1387 				break;
1388 
1389 			case O_PROTO:
1390 				/*
1391 				 * We do not allow an arg of 0 so the
1392 				 * check of "proto" only suffices.
1393 				 */
1394 				match = (proto == cmd->arg1);
1395 				break;
1396 
1397 			case O_IP_SRC:
1398 				match = is_ipv4 &&
1399 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1400 				    src_ip.s_addr);
1401 				break;
1402 
1403 			case O_IP_SRC_LOOKUP:
1404 			case O_IP_DST_LOOKUP:
1405 				if (is_ipv4) {
1406 				    uint32_t key =
1407 					(cmd->opcode == O_IP_DST_LOOKUP) ?
1408 					    dst_ip.s_addr : src_ip.s_addr;
1409 				    uint32_t v = 0;
1410 
1411 				    if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1412 					/* generic lookup. The key must be
1413 					 * in 32bit big-endian format.
1414 					 */
1415 					v = ((ipfw_insn_u32 *)cmd)->d[1];
1416 					if (v == 0)
1417 					    key = dst_ip.s_addr;
1418 					else if (v == 1)
1419 					    key = src_ip.s_addr;
1420 					else if (v == 6) /* dscp */
1421 					    key = (ip->ip_tos >> 2) & 0x3f;
1422 					else if (offset != 0)
1423 					    break;
1424 					else if (proto != IPPROTO_TCP &&
1425 						proto != IPPROTO_UDP)
1426 					    break;
1427 					else if (v == 2)
1428 					    key = htonl(dst_port);
1429 					else if (v == 3)
1430 					    key = htonl(src_port);
1431 					else if (v == 4 || v == 5) {
1432 					    check_uidgid(
1433 						(ipfw_insn_u32 *)cmd,
1434 						args, &ucred_lookup,
1435 #ifdef __FreeBSD__
1436 						&ucred_cache);
1437 					    if (v == 4 /* O_UID */)
1438 						key = ucred_cache->cr_uid;
1439 					    else if (v == 5 /* O_JAIL */)
1440 						key = ucred_cache->cr_prison->pr_id;
1441 #else /* !__FreeBSD__ */
1442 						(void *)&ucred_cache);
1443 					    if (v ==4 /* O_UID */)
1444 						key = ucred_cache.uid;
1445 					    else if (v == 5 /* O_JAIL */)
1446 						key = ucred_cache.xid;
1447 #endif /* !__FreeBSD__ */
1448 					    key = htonl(key);
1449 					} else
1450 					    break;
1451 				    }
1452 				    match = ipfw_lookup_table(chain,
1453 					cmd->arg1, key, &v);
1454 				    if (!match)
1455 					break;
1456 				    if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1457 					match =
1458 					    ((ipfw_insn_u32 *)cmd)->d[0] == v;
1459 				    else
1460 					tablearg = v;
1461 				} else if (is_ipv6) {
1462 					uint32_t v = 0;
1463 					void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1464 						&args->f_id.dst_ip6: &args->f_id.src_ip6;
1465 					match = ipfw_lookup_table_extended(chain,
1466 							cmd->arg1, pkey, &v,
1467 							IPFW_TABLE_CIDR);
1468 					if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1469 						match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1470 					if (match)
1471 						tablearg = v;
1472 				}
1473 				break;
1474 
1475 			case O_IP_SRC_MASK:
1476 			case O_IP_DST_MASK:
1477 				if (is_ipv4) {
1478 				    uint32_t a =
1479 					(cmd->opcode == O_IP_DST_MASK) ?
1480 					    dst_ip.s_addr : src_ip.s_addr;
1481 				    uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1482 				    int i = cmdlen-1;
1483 
1484 				    for (; !match && i>0; i-= 2, p+= 2)
1485 					match = (p[0] == (a & p[1]));
1486 				}
1487 				break;
1488 
1489 			case O_IP_SRC_ME:
1490 				if (is_ipv4) {
1491 					struct ifnet *tif;
1492 
1493 					INADDR_TO_IFP(src_ip, tif);
1494 					match = (tif != NULL);
1495 					break;
1496 				}
1497 #ifdef INET6
1498 				/* FALLTHROUGH */
1499 			case O_IP6_SRC_ME:
1500 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1501 #endif
1502 				break;
1503 
1504 			case O_IP_DST_SET:
1505 			case O_IP_SRC_SET:
1506 				if (is_ipv4) {
1507 					u_int32_t *d = (u_int32_t *)(cmd+1);
1508 					u_int32_t addr =
1509 					    cmd->opcode == O_IP_DST_SET ?
1510 						args->f_id.dst_ip :
1511 						args->f_id.src_ip;
1512 
1513 					    if (addr < d[0])
1514 						    break;
1515 					    addr -= d[0]; /* subtract base */
1516 					    match = (addr < cmd->arg1) &&
1517 						( d[ 1 + (addr>>5)] &
1518 						  (1<<(addr & 0x1f)) );
1519 				}
1520 				break;
1521 
1522 			case O_IP_DST:
1523 				match = is_ipv4 &&
1524 				    (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1525 				    dst_ip.s_addr);
1526 				break;
1527 
1528 			case O_IP_DST_ME:
1529 				if (is_ipv4) {
1530 					struct ifnet *tif;
1531 
1532 					INADDR_TO_IFP(dst_ip, tif);
1533 					match = (tif != NULL);
1534 					break;
1535 				}
1536 #ifdef INET6
1537 				/* FALLTHROUGH */
1538 			case O_IP6_DST_ME:
1539 				match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1540 #endif
1541 				break;
1542 
1543 
1544 			case O_IP_SRCPORT:
1545 			case O_IP_DSTPORT:
1546 				/*
1547 				 * offset == 0 && proto != 0 is enough
1548 				 * to guarantee that we have a
1549 				 * packet with port info.
1550 				 */
1551 				if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1552 				    && offset == 0) {
1553 					u_int16_t x =
1554 					    (cmd->opcode == O_IP_SRCPORT) ?
1555 						src_port : dst_port ;
1556 					u_int16_t *p =
1557 					    ((ipfw_insn_u16 *)cmd)->ports;
1558 					int i;
1559 
1560 					for (i = cmdlen - 1; !match && i>0;
1561 					    i--, p += 2)
1562 						match = (x>=p[0] && x<=p[1]);
1563 				}
1564 				break;
1565 
1566 			case O_ICMPTYPE:
1567 				match = (offset == 0 && proto==IPPROTO_ICMP &&
1568 				    icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1569 				break;
1570 
1571 #ifdef INET6
1572 			case O_ICMP6TYPE:
1573 				match = is_ipv6 && offset == 0 &&
1574 				    proto==IPPROTO_ICMPV6 &&
1575 				    icmp6type_match(
1576 					ICMP6(ulp)->icmp6_type,
1577 					(ipfw_insn_u32 *)cmd);
1578 				break;
1579 #endif /* INET6 */
1580 
1581 			case O_IPOPT:
1582 				match = (is_ipv4 &&
1583 				    ipopts_match(ip, cmd) );
1584 				break;
1585 
1586 			case O_IPVER:
1587 				match = (is_ipv4 &&
1588 				    cmd->arg1 == ip->ip_v);
1589 				break;
1590 
1591 			case O_IPID:
1592 			case O_IPLEN:
1593 			case O_IPTTL:
1594 				if (is_ipv4) {	/* only for IP packets */
1595 				    uint16_t x;
1596 				    uint16_t *p;
1597 				    int i;
1598 
1599 				    if (cmd->opcode == O_IPLEN)
1600 					x = iplen;
1601 				    else if (cmd->opcode == O_IPTTL)
1602 					x = ip->ip_ttl;
1603 				    else /* must be IPID */
1604 					x = ntohs(ip->ip_id);
1605 				    if (cmdlen == 1) {
1606 					match = (cmd->arg1 == x);
1607 					break;
1608 				    }
1609 				    /* otherwise we have ranges */
1610 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1611 				    i = cmdlen - 1;
1612 				    for (; !match && i>0; i--, p += 2)
1613 					match = (x >= p[0] && x <= p[1]);
1614 				}
1615 				break;
1616 
1617 			case O_IPPRECEDENCE:
1618 				match = (is_ipv4 &&
1619 				    (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1620 				break;
1621 
1622 			case O_IPTOS:
1623 				match = (is_ipv4 &&
1624 				    flags_match(cmd, ip->ip_tos));
1625 				break;
1626 
1627 			case O_TCPDATALEN:
1628 				if (proto == IPPROTO_TCP && offset == 0) {
1629 				    struct tcphdr *tcp;
1630 				    uint16_t x;
1631 				    uint16_t *p;
1632 				    int i;
1633 
1634 				    tcp = TCP(ulp);
1635 				    x = iplen -
1636 					((ip->ip_hl + tcp->th_off) << 2);
1637 				    if (cmdlen == 1) {
1638 					match = (cmd->arg1 == x);
1639 					break;
1640 				    }
1641 				    /* otherwise we have ranges */
1642 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1643 				    i = cmdlen - 1;
1644 				    for (; !match && i>0; i--, p += 2)
1645 					match = (x >= p[0] && x <= p[1]);
1646 				}
1647 				break;
1648 
1649 			case O_TCPFLAGS:
1650 				match = (proto == IPPROTO_TCP && offset == 0 &&
1651 				    flags_match(cmd, TCP(ulp)->th_flags));
1652 				break;
1653 
1654 			case O_TCPOPTS:
1655 				PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1656 				match = (proto == IPPROTO_TCP && offset == 0 &&
1657 				    tcpopts_match(TCP(ulp), cmd));
1658 				break;
1659 
1660 			case O_TCPSEQ:
1661 				match = (proto == IPPROTO_TCP && offset == 0 &&
1662 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1663 					TCP(ulp)->th_seq);
1664 				break;
1665 
1666 			case O_TCPACK:
1667 				match = (proto == IPPROTO_TCP && offset == 0 &&
1668 				    ((ipfw_insn_u32 *)cmd)->d[0] ==
1669 					TCP(ulp)->th_ack);
1670 				break;
1671 
1672 			case O_TCPWIN:
1673 				if (proto == IPPROTO_TCP && offset == 0) {
1674 				    uint16_t x;
1675 				    uint16_t *p;
1676 				    int i;
1677 
1678 				    x = ntohs(TCP(ulp)->th_win);
1679 				    if (cmdlen == 1) {
1680 					match = (cmd->arg1 == x);
1681 					break;
1682 				    }
1683 				    /* Otherwise we have ranges. */
1684 				    p = ((ipfw_insn_u16 *)cmd)->ports;
1685 				    i = cmdlen - 1;
1686 				    for (; !match && i > 0; i--, p += 2)
1687 					match = (x >= p[0] && x <= p[1]);
1688 				}
1689 				break;
1690 
1691 			case O_ESTAB:
1692 				/* reject packets which have SYN only */
1693 				/* XXX should i also check for TH_ACK ? */
1694 				match = (proto == IPPROTO_TCP && offset == 0 &&
1695 				    (TCP(ulp)->th_flags &
1696 				     (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1697 				break;
1698 
1699 			case O_ALTQ: {
1700 				struct pf_mtag *at;
1701 				struct m_tag *mtag;
1702 				ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1703 
1704 				/*
1705 				 * ALTQ uses mbuf tags from another
1706 				 * packet filtering system - pf(4).
1707 				 * We allocate a tag in its format
1708 				 * and fill it in, pretending to be pf(4).
1709 				 */
1710 				match = 1;
1711 				at = pf_find_mtag(m);
1712 				if (at != NULL && at->qid != 0)
1713 					break;
1714 				mtag = m_tag_get(PACKET_TAG_PF,
1715 				    sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
1716 				if (mtag == NULL) {
1717 					/*
1718 					 * Let the packet fall back to the
1719 					 * default ALTQ.
1720 					 */
1721 					break;
1722 				}
1723 				m_tag_prepend(m, mtag);
1724 				at = (struct pf_mtag *)(mtag + 1);
1725 				at->qid = altq->qid;
1726 				at->hdr = ip;
1727 				break;
1728 			}
1729 
1730 			case O_LOG:
1731 				ipfw_log(f, hlen, args, m,
1732 				    oif, offset | ip6f_mf, tablearg, ip);
1733 				match = 1;
1734 				break;
1735 
1736 			case O_PROB:
1737 				match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1738 				break;
1739 
1740 			case O_VERREVPATH:
1741 				/* Outgoing packets automatically pass/match */
1742 				match = ((oif != NULL) ||
1743 				    (m->m_pkthdr.rcvif == NULL) ||
1744 				    (
1745 #ifdef INET6
1746 				    is_ipv6 ?
1747 					verify_path6(&(args->f_id.src_ip6),
1748 					    m->m_pkthdr.rcvif, args->f_id.fib) :
1749 #endif
1750 				    verify_path(src_ip, m->m_pkthdr.rcvif,
1751 				        args->f_id.fib)));
1752 				break;
1753 
1754 			case O_VERSRCREACH:
1755 				/* Outgoing packets automatically pass/match */
1756 				match = (hlen > 0 && ((oif != NULL) ||
1757 #ifdef INET6
1758 				    is_ipv6 ?
1759 				        verify_path6(&(args->f_id.src_ip6),
1760 				            NULL, args->f_id.fib) :
1761 #endif
1762 				    verify_path(src_ip, NULL, args->f_id.fib)));
1763 				break;
1764 
1765 			case O_ANTISPOOF:
1766 				/* Outgoing packets automatically pass/match */
1767 				if (oif == NULL && hlen > 0 &&
1768 				    (  (is_ipv4 && in_localaddr(src_ip))
1769 #ifdef INET6
1770 				    || (is_ipv6 &&
1771 				        in6_localaddr(&(args->f_id.src_ip6)))
1772 #endif
1773 				    ))
1774 					match =
1775 #ifdef INET6
1776 					    is_ipv6 ? verify_path6(
1777 					        &(args->f_id.src_ip6),
1778 					        m->m_pkthdr.rcvif,
1779 						args->f_id.fib) :
1780 #endif
1781 					    verify_path(src_ip,
1782 					    	m->m_pkthdr.rcvif,
1783 					        args->f_id.fib);
1784 				else
1785 					match = 1;
1786 				break;
1787 
1788 			case O_IPSEC:
1789 #ifdef IPSEC
1790 				match = (m_tag_find(m,
1791 				    PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1792 #endif
1793 				/* otherwise no match */
1794 				break;
1795 
1796 #ifdef INET6
1797 			case O_IP6_SRC:
1798 				match = is_ipv6 &&
1799 				    IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1800 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1801 				break;
1802 
1803 			case O_IP6_DST:
1804 				match = is_ipv6 &&
1805 				IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1806 				    &((ipfw_insn_ip6 *)cmd)->addr6);
1807 				break;
1808 			case O_IP6_SRC_MASK:
1809 			case O_IP6_DST_MASK:
1810 				if (is_ipv6) {
1811 					int i = cmdlen - 1;
1812 					struct in6_addr p;
1813 					struct in6_addr *d =
1814 					    &((ipfw_insn_ip6 *)cmd)->addr6;
1815 
1816 					for (; !match && i > 0; d += 2,
1817 					    i -= F_INSN_SIZE(struct in6_addr)
1818 					    * 2) {
1819 						p = (cmd->opcode ==
1820 						    O_IP6_SRC_MASK) ?
1821 						    args->f_id.src_ip6:
1822 						    args->f_id.dst_ip6;
1823 						APPLY_MASK(&p, &d[1]);
1824 						match =
1825 						    IN6_ARE_ADDR_EQUAL(&d[0],
1826 						    &p);
1827 					}
1828 				}
1829 				break;
1830 
1831 			case O_FLOW6ID:
1832 				match = is_ipv6 &&
1833 				    flow6id_match(args->f_id.flow_id6,
1834 				    (ipfw_insn_u32 *) cmd);
1835 				break;
1836 
1837 			case O_EXT_HDR:
1838 				match = is_ipv6 &&
1839 				    (ext_hd & ((ipfw_insn *) cmd)->arg1);
1840 				break;
1841 
1842 			case O_IP6:
1843 				match = is_ipv6;
1844 				break;
1845 #endif
1846 
1847 			case O_IP4:
1848 				match = is_ipv4;
1849 				break;
1850 
1851 			case O_TAG: {
1852 				struct m_tag *mtag;
1853 				uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1854 				    tablearg : cmd->arg1;
1855 
1856 				/* Packet is already tagged with this tag? */
1857 				mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1858 
1859 				/* We have `untag' action when F_NOT flag is
1860 				 * present. And we must remove this mtag from
1861 				 * mbuf and reset `match' to zero (`match' will
1862 				 * be inversed later).
1863 				 * Otherwise we should allocate new mtag and
1864 				 * push it into mbuf.
1865 				 */
1866 				if (cmd->len & F_NOT) { /* `untag' action */
1867 					if (mtag != NULL)
1868 						m_tag_delete(m, mtag);
1869 					match = 0;
1870 				} else {
1871 					if (mtag == NULL) {
1872 						mtag = m_tag_alloc( MTAG_IPFW,
1873 						    tag, 0, M_NOWAIT);
1874 						if (mtag != NULL)
1875 							m_tag_prepend(m, mtag);
1876 					}
1877 					match = 1;
1878 				}
1879 				break;
1880 			}
1881 
1882 			case O_FIB: /* try match the specified fib */
1883 				if (args->f_id.fib == cmd->arg1)
1884 					match = 1;
1885 				break;
1886 
1887 			case O_SOCKARG:	{
1888 				struct inpcb *inp = args->inp;
1889 				struct inpcbinfo *pi;
1890 
1891 				if (is_ipv6) /* XXX can we remove this ? */
1892 					break;
1893 
1894 				if (proto == IPPROTO_TCP)
1895 					pi = &V_tcbinfo;
1896 				else if (proto == IPPROTO_UDP)
1897 					pi = &V_udbinfo;
1898 				else
1899 					break;
1900 
1901 				/*
1902 				 * XXXRW: so_user_cookie should almost
1903 				 * certainly be inp_user_cookie?
1904 				 */
1905 
1906 				/* For incomming packet, lookup up the
1907 				inpcb using the src/dest ip/port tuple */
1908 				if (inp == NULL) {
1909 					inp = in_pcblookup(pi,
1910 						src_ip, htons(src_port),
1911 						dst_ip, htons(dst_port),
1912 						INPLOOKUP_RLOCKPCB, NULL);
1913 					if (inp != NULL) {
1914 						tablearg =
1915 						    inp->inp_socket->so_user_cookie;
1916 						if (tablearg)
1917 							match = 1;
1918 						INP_RUNLOCK(inp);
1919 					}
1920 				} else {
1921 					if (inp->inp_socket) {
1922 						tablearg =
1923 						    inp->inp_socket->so_user_cookie;
1924 						if (tablearg)
1925 							match = 1;
1926 					}
1927 				}
1928 				break;
1929 			}
1930 
1931 			case O_TAGGED: {
1932 				struct m_tag *mtag;
1933 				uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1934 				    tablearg : cmd->arg1;
1935 
1936 				if (cmdlen == 1) {
1937 					match = m_tag_locate(m, MTAG_IPFW,
1938 					    tag, NULL) != NULL;
1939 					break;
1940 				}
1941 
1942 				/* we have ranges */
1943 				for (mtag = m_tag_first(m);
1944 				    mtag != NULL && !match;
1945 				    mtag = m_tag_next(m, mtag)) {
1946 					uint16_t *p;
1947 					int i;
1948 
1949 					if (mtag->m_tag_cookie != MTAG_IPFW)
1950 						continue;
1951 
1952 					p = ((ipfw_insn_u16 *)cmd)->ports;
1953 					i = cmdlen - 1;
1954 					for(; !match && i > 0; i--, p += 2)
1955 						match =
1956 						    mtag->m_tag_id >= p[0] &&
1957 						    mtag->m_tag_id <= p[1];
1958 				}
1959 				break;
1960 			}
1961 
1962 			/*
1963 			 * The second set of opcodes represents 'actions',
1964 			 * i.e. the terminal part of a rule once the packet
1965 			 * matches all previous patterns.
1966 			 * Typically there is only one action for each rule,
1967 			 * and the opcode is stored at the end of the rule
1968 			 * (but there are exceptions -- see below).
1969 			 *
1970 			 * In general, here we set retval and terminate the
1971 			 * outer loop (would be a 'break 3' in some language,
1972 			 * but we need to set l=0, done=1)
1973 			 *
1974 			 * Exceptions:
1975 			 * O_COUNT and O_SKIPTO actions:
1976 			 *   instead of terminating, we jump to the next rule
1977 			 *   (setting l=0), or to the SKIPTO target (setting
1978 			 *   f/f_len, cmd and l as needed), respectively.
1979 			 *
1980 			 * O_TAG, O_LOG and O_ALTQ action parameters:
1981 			 *   perform some action and set match = 1;
1982 			 *
1983 			 * O_LIMIT and O_KEEP_STATE: these opcodes are
1984 			 *   not real 'actions', and are stored right
1985 			 *   before the 'action' part of the rule.
1986 			 *   These opcodes try to install an entry in the
1987 			 *   state tables; if successful, we continue with
1988 			 *   the next opcode (match=1; break;), otherwise
1989 			 *   the packet must be dropped (set retval,
1990 			 *   break loops with l=0, done=1)
1991 			 *
1992 			 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1993 			 *   cause a lookup of the state table, and a jump
1994 			 *   to the 'action' part of the parent rule
1995 			 *   if an entry is found, or
1996 			 *   (CHECK_STATE only) a jump to the next rule if
1997 			 *   the entry is not found.
1998 			 *   The result of the lookup is cached so that
1999 			 *   further instances of these opcodes become NOPs.
2000 			 *   The jump to the next rule is done by setting
2001 			 *   l=0, cmdlen=0.
2002 			 */
2003 			case O_LIMIT:
2004 			case O_KEEP_STATE:
2005 				if (ipfw_install_state(f,
2006 				    (ipfw_insn_limit *)cmd, args, tablearg)) {
2007 					/* error or limit violation */
2008 					retval = IP_FW_DENY;
2009 					l = 0;	/* exit inner loop */
2010 					done = 1; /* exit outer loop */
2011 				}
2012 				match = 1;
2013 				break;
2014 
2015 			case O_PROBE_STATE:
2016 			case O_CHECK_STATE:
2017 				/*
2018 				 * dynamic rules are checked at the first
2019 				 * keep-state or check-state occurrence,
2020 				 * with the result being stored in dyn_dir.
2021 				 * The compiler introduces a PROBE_STATE
2022 				 * instruction for us when we have a
2023 				 * KEEP_STATE (because PROBE_STATE needs
2024 				 * to be run first).
2025 				 */
2026 				if (dyn_dir == MATCH_UNKNOWN &&
2027 				    (q = ipfw_lookup_dyn_rule(&args->f_id,
2028 				     &dyn_dir, proto == IPPROTO_TCP ?
2029 					TCP(ulp) : NULL))
2030 					!= NULL) {
2031 					/*
2032 					 * Found dynamic entry, update stats
2033 					 * and jump to the 'action' part of
2034 					 * the parent rule by setting
2035 					 * f, cmd, l and clearing cmdlen.
2036 					 */
2037 					q->pcnt++;
2038 					q->bcnt += pktlen;
2039 					/* XXX we would like to have f_pos
2040 					 * readily accessible in the dynamic
2041 				         * rule, instead of having to
2042 					 * lookup q->rule.
2043 					 */
2044 					f = q->rule;
2045 					f_pos = ipfw_find_rule(chain,
2046 						f->rulenum, f->id);
2047 					cmd = ACTION_PTR(f);
2048 					l = f->cmd_len - f->act_ofs;
2049 					ipfw_dyn_unlock();
2050 					cmdlen = 0;
2051 					match = 1;
2052 					break;
2053 				}
2054 				/*
2055 				 * Dynamic entry not found. If CHECK_STATE,
2056 				 * skip to next rule, if PROBE_STATE just
2057 				 * ignore and continue with next opcode.
2058 				 */
2059 				if (cmd->opcode == O_CHECK_STATE)
2060 					l = 0;	/* exit inner loop */
2061 				match = 1;
2062 				break;
2063 
2064 			case O_ACCEPT:
2065 				retval = 0;	/* accept */
2066 				l = 0;		/* exit inner loop */
2067 				done = 1;	/* exit outer loop */
2068 				break;
2069 
2070 			case O_PIPE:
2071 			case O_QUEUE:
2072 				set_match(args, f_pos, chain);
2073 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2074 					tablearg : cmd->arg1;
2075 				if (cmd->opcode == O_PIPE)
2076 					args->rule.info |= IPFW_IS_PIPE;
2077 				if (V_fw_one_pass)
2078 					args->rule.info |= IPFW_ONEPASS;
2079 				retval = IP_FW_DUMMYNET;
2080 				l = 0;          /* exit inner loop */
2081 				done = 1;       /* exit outer loop */
2082 				break;
2083 
2084 			case O_DIVERT:
2085 			case O_TEE:
2086 				if (args->eh) /* not on layer 2 */
2087 				    break;
2088 				/* otherwise this is terminal */
2089 				l = 0;		/* exit inner loop */
2090 				done = 1;	/* exit outer loop */
2091 				retval = (cmd->opcode == O_DIVERT) ?
2092 					IP_FW_DIVERT : IP_FW_TEE;
2093 				set_match(args, f_pos, chain);
2094 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2095 				    tablearg : cmd->arg1;
2096 				break;
2097 
2098 			case O_COUNT:
2099 				f->pcnt++;	/* update stats */
2100 				f->bcnt += pktlen;
2101 				f->timestamp = time_uptime;
2102 				l = 0;		/* exit inner loop */
2103 				break;
2104 
2105 			case O_SKIPTO:
2106 			    f->pcnt++;	/* update stats */
2107 			    f->bcnt += pktlen;
2108 			    f->timestamp = time_uptime;
2109 			    /* If possible use cached f_pos (in f->next_rule),
2110 			     * whose version is written in f->next_rule
2111 			     * (horrible hacks to avoid changing the ABI).
2112 			     */
2113 			    if (cmd->arg1 != IP_FW_TABLEARG &&
2114 				    (uintptr_t)f->x_next == chain->id) {
2115 				f_pos = (uintptr_t)f->next_rule;
2116 			    } else {
2117 				int i = (cmd->arg1 == IP_FW_TABLEARG) ?
2118 					tablearg : cmd->arg1;
2119 				/* make sure we do not jump backward */
2120 				if (i <= f->rulenum)
2121 				    i = f->rulenum + 1;
2122 				f_pos = ipfw_find_rule(chain, i, 0);
2123 				/* update the cache */
2124 				if (cmd->arg1 != IP_FW_TABLEARG) {
2125 				    f->next_rule =
2126 					(void *)(uintptr_t)f_pos;
2127 				    f->x_next =
2128 					(void *)(uintptr_t)chain->id;
2129 				}
2130 			    }
2131 			    /*
2132 			     * Skip disabled rules, and re-enter
2133 			     * the inner loop with the correct
2134 			     * f_pos, f, l and cmd.
2135 			     * Also clear cmdlen and skip_or
2136 			     */
2137 			    for (; f_pos < chain->n_rules - 1 &&
2138 				    (V_set_disable &
2139 				     (1 << chain->map[f_pos]->set));
2140 				    f_pos++)
2141 				;
2142 			    /* Re-enter the inner loop at the skipto rule. */
2143 			    f = chain->map[f_pos];
2144 			    l = f->cmd_len;
2145 			    cmd = f->cmd;
2146 			    match = 1;
2147 			    cmdlen = 0;
2148 			    skip_or = 0;
2149 			    continue;
2150 			    break;	/* not reached */
2151 
2152 			case O_CALLRETURN: {
2153 				/*
2154 				 * Implementation of `subroutine' call/return,
2155 				 * in the stack carried in an mbuf tag. This
2156 				 * is different from `skipto' in that any call
2157 				 * address is possible (`skipto' must prevent
2158 				 * backward jumps to avoid endless loops).
2159 				 * We have `return' action when F_NOT flag is
2160 				 * present. The `m_tag_id' field is used as
2161 				 * stack pointer.
2162 				 */
2163 				struct m_tag *mtag;
2164 				uint16_t jmpto, *stack;
2165 
2166 #define	IS_CALL		((cmd->len & F_NOT) == 0)
2167 #define	IS_RETURN	((cmd->len & F_NOT) != 0)
2168 				/*
2169 				 * Hand-rolled version of m_tag_locate() with
2170 				 * wildcard `type'.
2171 				 * If not already tagged, allocate new tag.
2172 				 */
2173 				mtag = m_tag_first(m);
2174 				while (mtag != NULL) {
2175 					if (mtag->m_tag_cookie ==
2176 					    MTAG_IPFW_CALL)
2177 						break;
2178 					mtag = m_tag_next(m, mtag);
2179 				}
2180 				if (mtag == NULL && IS_CALL) {
2181 					mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2182 					    IPFW_CALLSTACK_SIZE *
2183 					    sizeof(uint16_t), M_NOWAIT);
2184 					if (mtag != NULL)
2185 						m_tag_prepend(m, mtag);
2186 				}
2187 
2188 				/*
2189 				 * On error both `call' and `return' just
2190 				 * continue with next rule.
2191 				 */
2192 				if (IS_RETURN && (mtag == NULL ||
2193 				    mtag->m_tag_id == 0)) {
2194 					l = 0;		/* exit inner loop */
2195 					break;
2196 				}
2197 				if (IS_CALL && (mtag == NULL ||
2198 				    mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2199 					printf("ipfw: call stack error, "
2200 					    "go to next rule\n");
2201 					l = 0;		/* exit inner loop */
2202 					break;
2203 				}
2204 
2205 				f->pcnt++;	/* update stats */
2206 				f->bcnt += pktlen;
2207 				f->timestamp = time_uptime;
2208 				stack = (uint16_t *)(mtag + 1);
2209 
2210 				/*
2211 				 * The `call' action may use cached f_pos
2212 				 * (in f->next_rule), whose version is written
2213 				 * in f->next_rule.
2214 				 * The `return' action, however, doesn't have
2215 				 * fixed jump address in cmd->arg1 and can't use
2216 				 * cache.
2217 				 */
2218 				if (IS_CALL) {
2219 					stack[mtag->m_tag_id] = f->rulenum;
2220 					mtag->m_tag_id++;
2221 					if (cmd->arg1 != IP_FW_TABLEARG &&
2222 					    (uintptr_t)f->x_next == chain->id) {
2223 						f_pos = (uintptr_t)f->next_rule;
2224 					} else {
2225 						jmpto = (cmd->arg1 ==
2226 						    IP_FW_TABLEARG) ? tablearg:
2227 						    cmd->arg1;
2228 						f_pos = ipfw_find_rule(chain,
2229 						    jmpto, 0);
2230 						/* update the cache */
2231 						if (cmd->arg1 !=
2232 						    IP_FW_TABLEARG) {
2233 							f->next_rule =
2234 							    (void *)(uintptr_t)
2235 							    f_pos;
2236 							f->x_next =
2237 							    (void *)(uintptr_t)
2238 							    chain->id;
2239 						}
2240 					}
2241 				} else {	/* `return' action */
2242 					mtag->m_tag_id--;
2243 					jmpto = stack[mtag->m_tag_id] + 1;
2244 					f_pos = ipfw_find_rule(chain, jmpto, 0);
2245 				}
2246 
2247 				/*
2248 				 * Skip disabled rules, and re-enter
2249 				 * the inner loop with the correct
2250 				 * f_pos, f, l and cmd.
2251 				 * Also clear cmdlen and skip_or
2252 				 */
2253 				for (; f_pos < chain->n_rules - 1 &&
2254 				    (V_set_disable &
2255 				    (1 << chain->map[f_pos]->set)); f_pos++)
2256 					;
2257 				/* Re-enter the inner loop at the dest rule. */
2258 				f = chain->map[f_pos];
2259 				l = f->cmd_len;
2260 				cmd = f->cmd;
2261 				cmdlen = 0;
2262 				skip_or = 0;
2263 				continue;
2264 				break;	/* NOTREACHED */
2265 			}
2266 #undef IS_CALL
2267 #undef IS_RETURN
2268 
2269 			case O_REJECT:
2270 				/*
2271 				 * Drop the packet and send a reject notice
2272 				 * if the packet is not ICMP (or is an ICMP
2273 				 * query), and it is not multicast/broadcast.
2274 				 */
2275 				if (hlen > 0 && is_ipv4 && offset == 0 &&
2276 				    (proto != IPPROTO_ICMP ||
2277 				     is_icmp_query(ICMP(ulp))) &&
2278 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2279 				    !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2280 					send_reject(args, cmd->arg1, iplen, ip);
2281 					m = args->m;
2282 				}
2283 				/* FALLTHROUGH */
2284 #ifdef INET6
2285 			case O_UNREACH6:
2286 				if (hlen > 0 && is_ipv6 &&
2287 				    ((offset & IP6F_OFF_MASK) == 0) &&
2288 				    (proto != IPPROTO_ICMPV6 ||
2289 				     (is_icmp6_query(icmp6_type) == 1)) &&
2290 				    !(m->m_flags & (M_BCAST|M_MCAST)) &&
2291 				    !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2292 					send_reject6(
2293 					    args, cmd->arg1, hlen,
2294 					    (struct ip6_hdr *)ip);
2295 					m = args->m;
2296 				}
2297 				/* FALLTHROUGH */
2298 #endif
2299 			case O_DENY:
2300 				retval = IP_FW_DENY;
2301 				l = 0;		/* exit inner loop */
2302 				done = 1;	/* exit outer loop */
2303 				break;
2304 
2305 			case O_FORWARD_IP:
2306 				if (args->eh)	/* not valid on layer2 pkts */
2307 					break;
2308 				if (q == NULL || q->rule != f ||
2309 				    dyn_dir == MATCH_FORWARD) {
2310 				    struct sockaddr_in *sa;
2311 				    sa = &(((ipfw_insn_sa *)cmd)->sa);
2312 				    if (sa->sin_addr.s_addr == INADDR_ANY) {
2313 					bcopy(sa, &args->hopstore,
2314 							sizeof(*sa));
2315 					args->hopstore.sin_addr.s_addr =
2316 						    htonl(tablearg);
2317 					args->next_hop = &args->hopstore;
2318 				    } else {
2319 					args->next_hop = sa;
2320 				    }
2321 				}
2322 				retval = IP_FW_PASS;
2323 				l = 0;          /* exit inner loop */
2324 				done = 1;       /* exit outer loop */
2325 				break;
2326 
2327 #ifdef INET6
2328 			case O_FORWARD_IP6:
2329 				if (args->eh)	/* not valid on layer2 pkts */
2330 					break;
2331 				if (q == NULL || q->rule != f ||
2332 				    dyn_dir == MATCH_FORWARD) {
2333 					struct sockaddr_in6 *sin6;
2334 
2335 					sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2336 					args->next_hop6 = sin6;
2337 				}
2338 				retval = IP_FW_PASS;
2339 				l = 0;		/* exit inner loop */
2340 				done = 1;	/* exit outer loop */
2341 				break;
2342 #endif
2343 
2344 			case O_NETGRAPH:
2345 			case O_NGTEE:
2346 				set_match(args, f_pos, chain);
2347 				args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2348 					tablearg : cmd->arg1;
2349 				if (V_fw_one_pass)
2350 					args->rule.info |= IPFW_ONEPASS;
2351 				retval = (cmd->opcode == O_NETGRAPH) ?
2352 				    IP_FW_NETGRAPH : IP_FW_NGTEE;
2353 				l = 0;          /* exit inner loop */
2354 				done = 1;       /* exit outer loop */
2355 				break;
2356 
2357 			case O_SETFIB: {
2358 				uint32_t fib;
2359 
2360 				f->pcnt++;	/* update stats */
2361 				f->bcnt += pktlen;
2362 				f->timestamp = time_uptime;
2363 				fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg:
2364 				    cmd->arg1;
2365 				if (fib >= rt_numfibs)
2366 					fib = 0;
2367 				M_SETFIB(m, fib);
2368 				args->f_id.fib = fib;
2369 				l = 0;		/* exit inner loop */
2370 				break;
2371 		        }
2372 
2373 			case O_NAT:
2374  				if (!IPFW_NAT_LOADED) {
2375 				    retval = IP_FW_DENY;
2376 				} else {
2377 				    struct cfg_nat *t;
2378 				    int nat_id;
2379 
2380 				    set_match(args, f_pos, chain);
2381 				    /* Check if this is 'global' nat rule */
2382 				    if (cmd->arg1 == 0) {
2383 					    retval = ipfw_nat_ptr(args, NULL, m);
2384 					    l = 0;
2385 					    done = 1;
2386 					    break;
2387 				    }
2388 				    t = ((ipfw_insn_nat *)cmd)->nat;
2389 				    if (t == NULL) {
2390 					nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
2391 						tablearg : cmd->arg1;
2392 					t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2393 
2394 					if (t == NULL) {
2395 					    retval = IP_FW_DENY;
2396 					    l = 0;	/* exit inner loop */
2397 					    done = 1;	/* exit outer loop */
2398 					    break;
2399 					}
2400 					if (cmd->arg1 != IP_FW_TABLEARG)
2401 					    ((ipfw_insn_nat *)cmd)->nat = t;
2402 				    }
2403 				    retval = ipfw_nat_ptr(args, t, m);
2404 				}
2405 				l = 0;          /* exit inner loop */
2406 				done = 1;       /* exit outer loop */
2407 				break;
2408 
2409 			case O_REASS: {
2410 				int ip_off;
2411 
2412 				f->pcnt++;
2413 				f->bcnt += pktlen;
2414 				l = 0;	/* in any case exit inner loop */
2415 				ip_off = ntohs(ip->ip_off);
2416 
2417 				/* if not fragmented, go to next rule */
2418 				if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2419 				    break;
2420 
2421 				args->m = m = ip_reass(m);
2422 
2423 				/*
2424 				 * do IP header checksum fixup.
2425 				 */
2426 				if (m == NULL) { /* fragment got swallowed */
2427 				    retval = IP_FW_DENY;
2428 				} else { /* good, packet complete */
2429 				    int hlen;
2430 
2431 				    ip = mtod(m, struct ip *);
2432 				    hlen = ip->ip_hl << 2;
2433 				    ip->ip_sum = 0;
2434 				    if (hlen == sizeof(struct ip))
2435 					ip->ip_sum = in_cksum_hdr(ip);
2436 				    else
2437 					ip->ip_sum = in_cksum(m, hlen);
2438 				    retval = IP_FW_REASS;
2439 				    set_match(args, f_pos, chain);
2440 				}
2441 				done = 1;	/* exit outer loop */
2442 				break;
2443 			}
2444 
2445 			default:
2446 				panic("-- unknown opcode %d\n", cmd->opcode);
2447 			} /* end of switch() on opcodes */
2448 			/*
2449 			 * if we get here with l=0, then match is irrelevant.
2450 			 */
2451 
2452 			if (cmd->len & F_NOT)
2453 				match = !match;
2454 
2455 			if (match) {
2456 				if (cmd->len & F_OR)
2457 					skip_or = 1;
2458 			} else {
2459 				if (!(cmd->len & F_OR)) /* not an OR block, */
2460 					break;		/* try next rule    */
2461 			}
2462 
2463 		}	/* end of inner loop, scan opcodes */
2464 #undef PULLUP_LEN
2465 
2466 		if (done)
2467 			break;
2468 
2469 /* next_rule:; */	/* try next rule		*/
2470 
2471 	}		/* end of outer for, scan rules */
2472 
2473 	if (done) {
2474 		struct ip_fw *rule = chain->map[f_pos];
2475 		/* Update statistics */
2476 		rule->pcnt++;
2477 		rule->bcnt += pktlen;
2478 		rule->timestamp = time_uptime;
2479 	} else {
2480 		retval = IP_FW_DENY;
2481 		printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2482 	}
2483 	IPFW_RUNLOCK(chain);
2484 #ifdef __FreeBSD__
2485 	if (ucred_cache != NULL)
2486 		crfree(ucred_cache);
2487 #endif
2488 	return (retval);
2489 
2490 pullup_failed:
2491 	if (V_fw_verbose)
2492 		printf("ipfw: pullup failed\n");
2493 	return (IP_FW_DENY);
2494 }
2495 
2496 /*
2497  * Set maximum number of tables that can be used in given VNET ipfw instance.
2498  */
2499 #ifdef SYSCTL_NODE
2500 static int
2501 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2502 {
2503 	int error;
2504 	unsigned int ntables;
2505 
2506 	ntables = V_fw_tables_max;
2507 
2508 	error = sysctl_handle_int(oidp, &ntables, 0, req);
2509 	/* Read operation or some error */
2510 	if ((error != 0) || (req->newptr == NULL))
2511 		return (error);
2512 
2513 	return (ipfw_resize_tables(&V_layer3_chain, ntables));
2514 }
2515 #endif
2516 /*
2517  * Module and VNET glue
2518  */
2519 
2520 /*
2521  * Stuff that must be initialised only on boot or module load
2522  */
2523 static int
2524 ipfw_init(void)
2525 {
2526 	int error = 0;
2527 
2528 	ipfw_dyn_attach();
2529 	/*
2530  	 * Only print out this stuff the first time around,
2531 	 * when called from the sysinit code.
2532 	 */
2533 	printf("ipfw2 "
2534 #ifdef INET6
2535 		"(+ipv6) "
2536 #endif
2537 		"initialized, divert %s, nat %s, "
2538 		"default to %s, logging ",
2539 #ifdef IPDIVERT
2540 		"enabled",
2541 #else
2542 		"loadable",
2543 #endif
2544 #ifdef IPFIREWALL_NAT
2545 		"enabled",
2546 #else
2547 		"loadable",
2548 #endif
2549 		default_to_accept ? "accept" : "deny");
2550 
2551 	/*
2552 	 * Note: V_xxx variables can be accessed here but the vnet specific
2553 	 * initializer may not have been called yet for the VIMAGE case.
2554 	 * Tuneables will have been processed. We will print out values for
2555 	 * the default vnet.
2556 	 * XXX This should all be rationalized AFTER 8.0
2557 	 */
2558 	if (V_fw_verbose == 0)
2559 		printf("disabled\n");
2560 	else if (V_verbose_limit == 0)
2561 		printf("unlimited\n");
2562 	else
2563 		printf("limited to %d packets/entry by default\n",
2564 		    V_verbose_limit);
2565 
2566 	/* Check user-supplied table count for validness */
2567 	if (default_fw_tables > IPFW_TABLES_MAX)
2568 	  default_fw_tables = IPFW_TABLES_MAX;
2569 
2570 	ipfw_log_bpf(1); /* init */
2571 	return (error);
2572 }
2573 
2574 /*
2575  * Called for the removal of the last instance only on module unload.
2576  */
2577 static void
2578 ipfw_destroy(void)
2579 {
2580 
2581 	ipfw_log_bpf(0); /* uninit */
2582 	ipfw_dyn_detach();
2583 	printf("IP firewall unloaded\n");
2584 }
2585 
2586 /*
2587  * Stuff that must be initialized for every instance
2588  * (including the first of course).
2589  */
2590 static int
2591 vnet_ipfw_init(const void *unused)
2592 {
2593 	int error;
2594 	struct ip_fw *rule = NULL;
2595 	struct ip_fw_chain *chain;
2596 
2597 	chain = &V_layer3_chain;
2598 
2599 	/* First set up some values that are compile time options */
2600 	V_autoinc_step = 100;	/* bounded to 1..1000 in add_rule() */
2601 	V_fw_deny_unknown_exthdrs = 1;
2602 #ifdef IPFIREWALL_VERBOSE
2603 	V_fw_verbose = 1;
2604 #endif
2605 #ifdef IPFIREWALL_VERBOSE_LIMIT
2606 	V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2607 #endif
2608 #ifdef IPFIREWALL_NAT
2609 	LIST_INIT(&chain->nat);
2610 #endif
2611 
2612 	/* insert the default rule and create the initial map */
2613 	chain->n_rules = 1;
2614 	chain->static_len = sizeof(struct ip_fw);
2615 	chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2616 	if (chain->map)
2617 		rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2618 
2619 	/* Set initial number of tables */
2620 	V_fw_tables_max = default_fw_tables;
2621 	error = ipfw_init_tables(chain);
2622 	if (error) {
2623 		printf("ipfw2: setting up tables failed\n");
2624 		free(chain->map, M_IPFW);
2625 		free(rule, M_IPFW);
2626 		return (ENOSPC);
2627 	}
2628 
2629 	/* fill and insert the default rule */
2630 	rule->act_ofs = 0;
2631 	rule->rulenum = IPFW_DEFAULT_RULE;
2632 	rule->cmd_len = 1;
2633 	rule->set = RESVD_SET;
2634 	rule->cmd[0].len = 1;
2635 	rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2636 	chain->rules = chain->default_rule = chain->map[0] = rule;
2637 	chain->id = rule->id = 1;
2638 
2639 	IPFW_LOCK_INIT(chain);
2640 	ipfw_dyn_init();
2641 
2642 	/* First set up some values that are compile time options */
2643 	V_ipfw_vnet_ready = 1;		/* Open for business */
2644 
2645 	/*
2646 	 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
2647 	 * Even if the latter two fail we still keep the module alive
2648 	 * because the sockopt and layer2 paths are still useful.
2649 	 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2650 	 * so we can ignore the exact return value and just set a flag.
2651 	 *
2652 	 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2653 	 * changes in the underlying (per-vnet) variables trigger
2654 	 * immediate hook()/unhook() calls.
2655 	 * In layer2 we have the same behaviour, except that V_ether_ipfw
2656 	 * is checked on each packet because there are no pfil hooks.
2657 	 */
2658 	V_ip_fw_ctl_ptr = ipfw_ctl;
2659 	error = ipfw_attach_hooks(1);
2660 	return (error);
2661 }
2662 
2663 /*
2664  * Called for the removal of each instance.
2665  */
2666 static int
2667 vnet_ipfw_uninit(const void *unused)
2668 {
2669 	struct ip_fw *reap, *rule;
2670 	struct ip_fw_chain *chain = &V_layer3_chain;
2671 	int i;
2672 
2673 	V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2674 	/*
2675 	 * disconnect from ipv4, ipv6, layer2 and sockopt.
2676 	 * Then grab, release and grab again the WLOCK so we make
2677 	 * sure the update is propagated and nobody will be in.
2678 	 */
2679 	(void)ipfw_attach_hooks(0 /* detach */);
2680 	V_ip_fw_ctl_ptr = NULL;
2681 	IPFW_UH_WLOCK(chain);
2682 	IPFW_UH_WUNLOCK(chain);
2683 	IPFW_UH_WLOCK(chain);
2684 
2685 	IPFW_WLOCK(chain);
2686 	ipfw_dyn_uninit(0);	/* run the callout_drain */
2687 	IPFW_WUNLOCK(chain);
2688 
2689 	ipfw_destroy_tables(chain);
2690 	reap = NULL;
2691 	IPFW_WLOCK(chain);
2692 	for (i = 0; i < chain->n_rules; i++) {
2693 		rule = chain->map[i];
2694 		rule->x_next = reap;
2695 		reap = rule;
2696 	}
2697 	if (chain->map)
2698 		free(chain->map, M_IPFW);
2699 	IPFW_WUNLOCK(chain);
2700 	IPFW_UH_WUNLOCK(chain);
2701 	if (reap != NULL)
2702 		ipfw_reap_rules(reap);
2703 	IPFW_LOCK_DESTROY(chain);
2704 	ipfw_dyn_uninit(1);	/* free the remaining parts */
2705 	return 0;
2706 }
2707 
2708 /*
2709  * Module event handler.
2710  * In general we have the choice of handling most of these events by the
2711  * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2712  * use the SYSINIT handlers as they are more capable of expressing the
2713  * flow of control during module and vnet operations, so this is just
2714  * a skeleton. Note there is no SYSINIT equivalent of the module
2715  * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2716  */
2717 static int
2718 ipfw_modevent(module_t mod, int type, void *unused)
2719 {
2720 	int err = 0;
2721 
2722 	switch (type) {
2723 	case MOD_LOAD:
2724 		/* Called once at module load or
2725 	 	 * system boot if compiled in. */
2726 		break;
2727 	case MOD_QUIESCE:
2728 		/* Called before unload. May veto unloading. */
2729 		break;
2730 	case MOD_UNLOAD:
2731 		/* Called during unload. */
2732 		break;
2733 	case MOD_SHUTDOWN:
2734 		/* Called during system shutdown. */
2735 		break;
2736 	default:
2737 		err = EOPNOTSUPP;
2738 		break;
2739 	}
2740 	return err;
2741 }
2742 
2743 static moduledata_t ipfwmod = {
2744 	"ipfw",
2745 	ipfw_modevent,
2746 	0
2747 };
2748 
2749 /* Define startup order. */
2750 #define	IPFW_SI_SUB_FIREWALL	SI_SUB_PROTO_IFATTACHDOMAIN
2751 #define	IPFW_MODEVENT_ORDER	(SI_ORDER_ANY - 255) /* On boot slot in here. */
2752 #define	IPFW_MODULE_ORDER	(IPFW_MODEVENT_ORDER + 1) /* A little later. */
2753 #define	IPFW_VNET_ORDER		(IPFW_MODEVENT_ORDER + 2) /* Later still. */
2754 
2755 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2756 MODULE_VERSION(ipfw, 2);
2757 /* should declare some dependencies here */
2758 
2759 /*
2760  * Starting up. Done in order after ipfwmod() has been called.
2761  * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2762  */
2763 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2764 	    ipfw_init, NULL);
2765 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2766 	    vnet_ipfw_init, NULL);
2767 
2768 /*
2769  * Closing up shop. These are done in REVERSE ORDER, but still
2770  * after ipfwmod() has been called. Not called on reboot.
2771  * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2772  * or when the module is unloaded.
2773  */
2774 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2775 	    ipfw_destroy, NULL);
2776 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2777 	    vnet_ipfw_uninit, NULL);
2778 /* end of file */
2779